Lu Cheng, Mandal Mrinal
Department of Electrical and Computer Engineering, University of Alberta, Edmonton, Alberta, CanadaT6G 2V4.
Annu Int Conf IEEE Eng Med Biol Soc. 2012;2012:5355-9. doi: 10.1109/EMBC.2012.6347204.
In the diagnosis of skin melanoma by analyzing histopathological images, the segmentation of the epidermis area is an important step. This paper proposes a computer-aided technique for segmentation and analysis of the epidermis area in the whole slide skin histopathological images. Before the segmentation technique is employed, a monochromatic color channel that provides a good discriminant information between the epidermis and dermis areas is determined. In order to reduce the processing time and perform the analysis efficiently, we employ multi-resolution image analysis in the proposed segmentation technique. At first, a low resolution whole slide image is generated. We then segment the low resolution image using a global threshold method and shape analysis. Based on the segmented epidermis area, the layout of epidermis is determined and the high resolution image tiles of epidermis are generated for further manual or automated analysis. Experimental results on 16 different whole slide skin images show that the proposed technique provides a superior performance, about 92% sensitivity rate, 93% precision and 97% specificity rate.
在通过分析组织病理学图像诊断皮肤黑色素瘤时，表皮区域的分割是重要的一步。本文提出了一种计算机辅助技术，用于在全切片皮肤组织病理学图像中分割和分析表皮区域。在采用分割技术之前，确定一个能在表皮和真皮区域之间提供良好判别信息的单色颜色通道。为了减少处理时间并高效地进行分析，我们在所提出的分割技术中采用多分辨率图像分析。首先，生成低分辨率的全切片图像。然后，我们使用全局阈值方法和形状分析对低分辨率图像进行分割。基于分割出的表皮区域，确定表皮的布局，并生成表皮的高分辨率图像块以进行进一步的手动或自动分析。对16张不同的全切片皮肤图像的实验结果表明，所提出的技术具有卓越的性能，灵敏度约为92%，精确率为93%，特异度为97%。
